# A decade-long shift in use of energy devices for BABA robotic thyroidectomy: automated video analysis by deep learning

## TL;DR

This study uses deep learning to analyze a decade of surgical videos and finds a shift from one energy device to another in robotic thyroidectomy procedures.

## Contribution

The study introduces automated video analysis using deep learning to track changes in surgical device usage over time.

## Key findings

- Use of Harmonic ACE curved shears decreased from 72.3% in 2013 to 20.7% in 2021.
- Use of the Permanent cautery hook increased from 27.7% in 2013 to 79.3% in 2021.

## Abstract

Bilateral axillo-breast approach robotic thyroidectomy (BABA RT) is a well-established minimally invasive surgical option; however, shifts in preferences regarding the energy devices used for the procedure have not been analyzed quantitatively. This study retrospectively reviewed 81 BABA-RT videos from 2013 to 2021, and a YOLOv5 deep learning model was used to analyze the use of four energy devices: Harmonic ACE curved shears, Permanent cautery hook, ProGrasp forceps, and Maryland bipolar forceps. The data revealed a significant temporal shift in the type of device used. Use of Harmonic ACE curved shears fell from 72.3% in 2013 to 20.7% in 2021. By contrast, use of the Permanent cautery hook increased from 27.7% to 79.3% during the same period. There was no change in use of the other two devices. This study shows that deep learning is a tool that can be used to track surgical device usage based on video evidence. The findings indicate a gradual shift in the device chosen for BABA RT, from Harmonic ACE curved shears to a Permanent cautery hook, over the past decade, and reflect evolving surgical preferences and techniques.

The online version contains supplementary material available at 10.1007/s11701-025-03113-7.